Cause-Specific Association Measures for Multivariate Competing Risks Data

• Activity Number: 458
• Type: Topic Contributed
• Date/Time: Wednesday, August 4, 2010 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Yu Cheng*+ and Hao Wang
• Companies: University of Pittsburgh and University of Pittsburgh
• Keywords: Cause-specific hazard functions ; Cross hazard ratio ; Multivariate competing risks

Abstract

We propose a pseudo-likelihood estimator for an extended cause-specific cross hazard ratio (Bandeen-Roche & Liang 2002) from bivariate competing risks data to more complicated family structures, e.g., exchangeable sibship data and multiple mother-child pairs. We also adapt an equivalent association measure based on bivariate hazards functions (Cheng & Fine 2008) and develop a plug-in estimator. This greatly broadens their applications. The asymptotic properties of the two estimators are established by using empirical processes techniques and practical performances are compared with that of a U-statistic estimator (Cheng & Fine 2009) by simulation studies. The practical utility of the three approaches is illustrated in an analysis of the Cache County Study of Dementia.